From: Torsten Bögershausen <tboegi@web.de>
Date: Thu, 10 Mar 2016 07:42:35 GMT
Subject: Re: [PATCH] rebase -p: avoid grep on potentailly non-ASCII data
Message-ID: <56E1256B.9030308@web.de>
In-Reply-To: <xmqqvb4vgzxs.fsf@gitster.mtv.corp.google.com>

On 09.03.16 21:26, Junio C Hamano wrote:
> Anders Kaseorg <andersk@mit.edu> writes:
[]
>  sane_grep () {
> -	GREP_OPTIONS= LC_ALL=C grep "$@"
> +	GREP_OPTIONS= LC_ALL=C grep @@SANE_TEXT_GREP@@ "$@"
>  }
>  
>  sane_egrep () {
> -	GREP_OPTIONS= LC_ALL=C egrep "$@"
> +	GREP_OPTIONS= LC_ALL=C egrep @@SANE_TEXT_GREP@@ "$@"
>  }
>

I always wondered why we do LC_ALL=C.
Isn't that begging for trouble, when we feed UTF-8, ISO-8895-1
or other stuff into a program and say LC_ALL=C at the same time ?

On my Debian Linux system I have
LANG=en_US.UTF-8

and

$ locale -a
C
C.UTF-8
en_US.utf8
POSIX
--------------

Mac OS has LANG unset, and reports
locale -a
en_US
en_US.ISO8859-1
en_US.ISO8859-15
en_US.US-ASCII
en_US.UTF-8
#(and a lot more )
C
POSIX

-----
My Centos has 
LANG=en_US.UTF-8

and reports e.g.
en_US
en_US.iso88591
en_US.iso885915
en_US.utf8
(And many more)

In t0204 we have
    LANGUAGE=is LC_ALL="$is_IS_locale" git init repo >actual &&
which is based on
	# is_IS.UTF-8 on Solaris and FreeBSD, is_IS.utf8 on Debian
	is_IS_locale=$(locale -a 2>/dev/null |
in 
lib-gettext.sh

Is there something we can steal here ?


http://pubs.opengroup.org/onlinepubs/007908799/xbd/envvar.html